Huijuan Yang

Bayesian Learning for Spatial Filtering in an EEG-Based Brain–Computer Interface

Spatial filtering for EEG feature extraction and classification is an important tool in brain-computer interface. However, there is generally no established theory that links spatial filtering directly to Bayes classification error. To address this issue, this paper proposes and studies a Bayesian analysis theory for spatial filtering in relation to Bayes error. Following the maximum entropy principle, we introduce a gamma probability model for describing single-trial EEG power features. We then formulate and analyze the theoretical relationship between Bayes classification error and the so-called Rayleigh quotient, which is a function of spatial filters and basically measures the ratio in power features between two classes. This paper also reports our extensive study that examines the theory and its use in classification, using three publicly available EEG data sets and state-of-the-art spatial filtering techniques and various classifiers. Specifically, we validate the positive relationship between Bayes error and Rayleigh quotient in real EEG power features. Finally, we demonstrate that the Bayes error can be practically reduced by applying a new spatial filter with lower Rayleigh quotient.

Dynamic initiation and dual-tree complex wavelet feature-based classification of motor imagery of swallow EEG signals

The use of motor imagery-based brain computer interface has recently been shown to have potential for rehabilitation. This paper proposes a novel scheme to detect motor imagery of swallow from electroencephalography (EEG) signals for dysphagia rehabilitation. The proposed scheme extracts features from the coefficients of dual-tree complex wavelet transform (DT-CWT). A novel sliding window-based peak localization scheme is proposed to dynamically locate the initiation of tongue movement from Electromyography (EMG) signal. Subsequently, effective time segments are extracted from EEG signal for classification based on the detected dynamic initiation location. Comparisons are made between our proposed scheme with that of the three existing approaches. The results based on six healthy subjects show that an increase in averaged accuracy of 9.95% is achieved. Further, an increase in averaged accuracy of 8.02% is resulted comparing our proposed scheme by using and not using the dynamic initiation to extract the time segments. Classification results using EMG data confirm that our results are not due to movements artifacts. Statistical tests with 95% confidence to estimate the accuracy on the respective action at chance level show that five out of six subjects performed above chance level for our proposed dynamic initiation and wavelet feature-based approach.

On the use of convolutional neural networks and augmented CSP features for multi-class motor imagery of EEG signals classification.

Learning the deep structures and unknown correlations is important for the detection of motor imagery of EEG signals (MI-EEG). This study investigates the use of convolutional neural networks (CNNs) for the classification of multi-class MI-EEG signals. Augmented common spatial pattern (ACSP) features are generated based on pair-wise projection matrices, which covers various frequency ranges. We propose a frequency complementary feature map selection (FCMS) scheme by constraining the dependency among frequency bands. Experiments are conducted on BCI competition IV dataset IIa with 9 subjects. Averaged cross-validation accuracy of 68.45% and 69.27% is achieved for FCMS and all feature maps, respectively, which is significantly higher (4.53% and 5.34%) than random map selection and higher (1.44% and 2.26%) than filter-bank CSP (FBCSP). The results demonstrate that the CNNs are capable of learning discriminant, deep structure features for EEG classification without relying on the handcrafted features.

Detection of motor imagery of swallow EEG signals based on the dual-tree complex wavelet transform and adaptive model selection

OBJECTIVE Detection of motor imagery of hand/arm has been extensively studied for stroke rehabilitation. This paper firstly investigates the detection of motor imagery of swallow (MI-SW) and motor imagery of tongue protrusion (MI-Ton) in an attempt to find a novel solution for post-stroke dysphagia rehabilitation. Detection of MI-SW from a simple yet relevant modality such as MI-Ton is then investigated, motivated by the similarity in activation patterns between tongue movements and swallowing and there being fewer movement artifacts in performing tongue movements compared to swallowing. APPROACH Novel features were extracted based on the coefficients of the dual-tree complex wavelet transform to build multiple training models for detecting MI-SW. The session-to-session classification accuracy was boosted by adaptively selecting the training model to maximize the ratio of between-classes distances versus within-class distances, using features of training and evaluation data. MAIN RESULTS Our proposed method yielded averaged cross-validation (CV) classification accuracies of 70.89% and 73.79% for MI-SW and MI-Ton for ten healthy subjects, which are significantly better than the results from existing methods. In addition, averaged CV accuracies of 66.40% and 70.24% for MI-SW and MI-Ton were obtained for one stroke patient, demonstrating the detectability of MI-SW and MI-Ton from the idle state. Furthermore, averaged session-to-session classification accuracies of 72.08% and 70% were achieved for ten healthy subjects and one stroke patient using the MI-Ton model. SIGNIFICANCE These results and the subjectwise strong correlations in classification accuracies between MI-SW and MI-Ton demonstrated the feasibility of detecting MI-SW from MI-Ton models.

Maximum dependency and minimum redundancy-based channel selection for motor imagery of walking EEG signal detection

This paper proposes a novel method to detect motor imagery of walking for the rehabilitation of stroke patients using the Laplacian derivatives (LAD) of power averaged across frequency bands as the feature. We propose to select the most correlated channels by jointly considering the mutual information between the LAD power features of the channels and the class labels, and the redundancy between the LAD power features of the channel with that of the selected channels. Experiments are conducted on the EEG data collected for 11 healthy subjects using proposed method and compared with existing methods. The results show that the proposed method yielded an average classification accuracy of 67.19% by selecting as few as 4 LAD channels. An improved result of 71.45% and 73.23% are achieved by selecting 10 and 22 LAD channels, respectively. Comparison results revealed significantly superior performance of our proposed method compared to that obtained using common spatial pattern and filter bank with power features. Most importantly, our proposed method achieves significant better accuracy for poor BCI performers compared to existing methods. Thus, the results demonstrated the potential of using the proposed method for detecting motor imagery of walking for the rehabilitation of stroke patients.

Selection of effective EEG channels in brain computer interfaces based on inconsistencies of classifiers

This paper proposed a novel method to select the effective Electroencephalography (EEG) channels for the motor imagery tasks based on the inconsistencies from multiple classifiers. The inconsistency criterion for channel selection was designed based on the fluctuation of the classification accuracies among different classifiers when the noisy channels were included. These noisy channels were then identified and removed till a required number of channels was selected or a predefined classification accuracy with reference to baseline was obtained. Experiments conducted on a data set of 13 healthy subjects performing hand grasping and idle revealed that the EEG channels from the motor area were most frequently selected. Furthermore, the mean increases of 4.07%, 3.10% and 1.77% of the averaged accuracies in comparison with the four existing channel selection methods were achieved for the non-feedback, feedback and calibration sessions, respectively, by selecting as low as seven channels. These results further validated the effectiveness of our proposed method.

Feature consistency-based model adaptation in session-to-session classification: A study using motor imagery of swallow EEG signals

The performance degradation for session to session classification in brain computer interface is a critical problem. This paper proposes a novel method for model adaptation based on motor imagery of swallow EEG signal for dysphagia rehabilitation. A small amount of calibration testing data is utilized to select the model catering for test data. The features of the training and calibration testing data are firstly clustered and each cluster is labeled by the dominant label of the training data. The cluster with the minimum impurity is selected and the number of features consistent with the cluster label are calculated for both training and calibration testing data. Finally, the training model with the maximum number of consistent features is selected. Experiments conducted on motor imagery of swallow EEG data achieved an average accuracy of 74.29% and 72.64% with model adaptation for Laplacian derivates of power features and wavelet features, respectively. Further, an average accuracy increase of 2.9% is achieved with model adaptation using wavelet features, in comparison with that achieved without model adaptation, which is significant at 5% significance level as demonstrated in the statistical test.

Seizure detection based on spatiotemporal correlation and frequency regularity of scalp EEG

In this paper, a robust seizure detection system using scalp EEG signal is presented. Two most important and obvious characteristics of seizure EEG, signal variance, and frequency synchronization are carefully chosen as seizure detection indexes. To extract the representation of EEG variance, a spatiotemporal correlation structure is constructed based on space-delay covariance matrices with multi-scale temporal delay. The frequency synchronization of EEG is represented by a regularity index derived from wavelet packet transform. The extracted representations are combined to form a high-dimensional feature vector with redundant information. In order to reduce the redundancy, feature selection is performed using mutual information (MI) based on best individual features. The optimized set of features form a more compact feature vector for each 2-s epoch of multi-channel EEG. Feature vectors are then classified into ictal or interictal class using a linear support vector machine (SVM). To evaluate the proposed seizure detection system, unbiased leave-one-session-out cross-validation using clinical routine EEG from 7 patients are performed in experiments. The proposed method obtains average accuracy of 91.44% and average latency of 6.82 s, which outperforms other 7 commonly used methods. It is also demonstrated that the performance of our method is more robust since the standard deviation of results among patients is smaller than other methods.

Prediction and detection of seizures from simultaneous thalamic and scalp electroencephalography recordings

OBJECTIVE The authors explored the feasibility of seizure detection and prediction using signals recorded from the anterior thalamic nucleus, a major target for deep brain stimulation (DBS) in the treatment of epilepsy. METHODS Using data from 5 patients (13 seizures in total), the authors performed a feasibility study and analyzed the performance of a seizure prediction and detection algorithm applied to simultaneously acquired scalp and thalamic electroencephalography (EEG). The thalamic signal was obtained from DBS electrodes. The applied algorithm used the similarity index as a nonlinear measure for seizure identification, with patient-specific channel and threshold selection. Receiver operating characteristic (ROC) curves were calculated using data from all patients and channels to compare the performance between DBS and EEG recordings. RESULTS Thalamic DBS recordings were associated with a mean prediction rate of 84%, detection rate of 97%, and false-alarm rate of 0.79/hr. In comparison, scalp EEG recordings were associated with a mean prediction rate of 71%, detection rate of 100%, and false-alarm rate of 1.01/hr. From the ROC curves, when considering all channels, DBS outperformed EEG for both detection and prediction of seizures. CONCLUSIONS This is the first study to compare automated seizure detection and prediction from simultaneous thalamic and scalp EEG recordings. The authors have demonstrated that signals recorded from DBS leads are more robust than EEG recordings and can be used to predict and detect seizures. These results indicate feasibility for future designs of closed-loop anterior nucleus DBS systems for the treatment of epilepsy.

Neural and cortical analysis of swallowing and detection of motor imagery of swallow for dysphagia rehabilitation-A review.

Swallowing is an essential function in our daily life; nevertheless, stroke or other neurodegenerative diseases can cause the malfunction of swallowing function, ie, dysphagia. The objectives of this review are to understand the neural and cortical basis of swallowing and tongue, and review the latest techniques on the detection of motor imagery of swallow (MI-SW) and motor imagery of tongue movements (MI-TM), so that a practical system can be developed for the rehabilitation of poststroke dysphagia patients. Specifically, we firstly describe the swallowing process and how the swallowing function is assessed clinically. Secondly, we review the techniques that performed the neural and cortical analysis of swallowing and tongue based on different modalities such as functional magnetic resonance imaging, positron emission tomography, near-infrared spectroscopy (NIRS), and magnetoencephalography. Thirdly, we review the techniques that performed detection and analysis of MI-SW and MI-TM for dysphagia stroke rehabilitation based on electroencephalography (EEG) and NIRS. Finally, discussions on the advantages and limitations of the studies are presented; an example system and future research directions for the rehabilitation of stroke dysphagia patients are suggested.