Cuntai Guan

Temporal classification of multichannel near-infrared spectroscopy signals of motor imagery for developing a brain–computer interface

There has been an increase in research interest for brain-computer interface (BCI) technology as an alternate mode of communication and environmental control for the disabled, such as patients suffering from amyotrophic lateral sclerosis (ALS), brainstem stroke and spinal cord injury. Disabled patients with appropriate physical care and cognitive ability to communicate with their social environment continue to live with a reasonable quality of life over extended periods of time. Near-infrared spectroscopy is a non-invasive technique which utilizes light in the near-infrared range (700 to 1000 nm) to determine cerebral oxygenation, blood flow and metabolic status of localized regions of the brain. In this paper, we describe a study conducted to test the feasibility of using multichannel NIRS in the development of a BCI. We used a continuous wave 20-channel NIRS system over the motor cortex of 5 healthy volunteers to measure oxygenated and deoxygenated hemoglobin changes during left-hand and right-hand motor imagery. We present results of signal analysis indicating that there exist distinct patterns of hemodynamic responses which could be utilized in a pattern classifier towards developing a BCI. We applied two different pattern recognition algorithms separately, Support Vector Machines (SVM) and Hidden Markov Model (HMM), to classify the data offline. SVM classified left-hand imagery from right-hand imagery with an average accuracy of 73% for all volunteers, while HMM performed better with an average accuracy of 89%. Our results indicate potential application of NIRS in the development of BCIs. We also discuss here future extension of our system to develop a word speller application based on a cursor control paradigm incorporating online pattern classification of single-trial NIRS data.

Regularizing Common Spatial Patterns to Improve BCI Designs: Unified Theory and New Algorithms

One of the most popular feature extraction algorithms for brain-computer interfaces (BCI) is common spatial patterns (CSPs). Despite its known efficiency and widespread use, CSP is also known to be very sensitive to noise and prone to overfitting. To address this issue, it has been recently proposed to regularize CSP. In this paper, we present a simple and unifying theoretical framework to design such a regularized CSP (RCSP). We then present a review of existing RCSP algorithms and describe how to cast them in this framework. We also propose four new RCSP algorithms. Finally, we compare the performances of 11 different RCSP (including the four new ones and the original CSP), on electroencephalography data from 17 subjects, from BCI competition datasets. Results showed that the best RCSP methods can outperform CSP by nearly 10% in median classification accuracy and lead to more neurophysiologically relevant spatial filters. They also enable us to perform efficient subject-to-subject transfer. Overall, the best RCSP algorithms were CSP with Tikhonov regularization and weighted Tikhonov regularization, both proposed in this paper.

Filter Bank Common Spatial Pattern (FBCSP) in Brain-Computer Interface

In motor imagery-based brain computer interfaces (BCI), discriminative patterns can be extracted from the electroencephalogram (EEG) using the common spatial pattern (CSP) algorithm. However, the performance of this spatial filter depends on the operational frequency band of the EEG. Thus, setting a broad frequency range, or manually selecting a subject-specific frequency range, are commonly used with the CSP algorithm. To address this problem, this paper proposes a novel filter bank common spatial pattern (FBCSP) to perform autonomous selection of key temporal-spatial discriminative EEG characteristics. After the EEG measurements have been bandpass-filtered into multiple frequency bands, CSP features are extracted from each of these bands. A feature selection algorithm is then used to automatically select discriminative pairs of frequency bands and corresponding CSP features. A classification algorithm is subsequently used to classify the CSP features. A study is conducted to assess the performance of a selection of feature selection and classification algorithms for use with the FBCSP. Extensive experimental results are presented on a publicly available dataset as well as data collected from healthy subjects and unilaterally paralyzed stroke patients. The results show that FBCSP, using a particular combination feature selection and classification algorithm, yields relatively higher cross-validation accuracies compared to prevailing approaches.

A Brain Controlled Wheelchair to Navigate in Familiar Environments

While brain-computer interfaces (BCIs) can provide communication to people who are locked-in, they suffer from a very low information transfer rate. Further, using a BCI requires a concentration effort and using it continuously can be tiring. The brain controlled wheelchair (BCW) described in this paper aims at providing mobility to BCI users despite these limitations, in a safe and efficient way. Using a slow but reliable P300 based BCI, the user selects a destination amongst a list of predefined locations. While the wheelchair moves on virtual guiding paths ensuring smooth, safe, and predictable trajectories, the user can stop the wheelchair by using a faster BCI. Experiments with nondisabled subjects demonstrated the efficiency of this strategy. Brain control was not affected when the wheelchair was in motion, and the BCW enabled the users to move to various locations in less time and with significantly less control effort than other control strategies proposed in the literature.

Robust classification of EEG signal for brain-computer interface

We report the implementation of a text input application (speller) based on the P300 event related potential. We obtain high accuracies by using an SVM classifier and a novel feature. These techniques enable us to maintain fast performance without sacrificing the accuracy, thus making the speller usable in an online mode. In order to further improve the usability, we perform various studies on the data with a view to minimizing the training time required. We present data collected from nine healthy subjects, along with the high accuracies (of the order of 95% or more) measured online. We show that the training time can be further reduced by a factor of two from its current value of about 20 min. High accuracy, fast learning, and online performance make this P300 speller a potential communication tool for severely disabled individuals, who have lost all other means of communication and are otherwise cut off from the world, provided their disability does not interfere with the performance of the speller.

Filter Bank Common Spatial Pattern Algorithm on BCI Competition IV Datasets 2a and 2b

The Common Spatial Pattern (CSP) algorithm is an effective and popular method for classifying 2-class motor imagery electroencephalogram (EEG) data, but its effectiveness depends on the subject-specific frequency band. This paper presents the Filter Bank Common Spatial Pattern (FBCSP) algorithm to optimize the subject-specific frequency band for CSP on Datasets 2a and 2b of the Brain-Computer Interface (BCI) Competition IV. Dataset 2a comprised 4 classes of 22 channels EEG data from 9 subjects, and Dataset 2b comprised 2 classes of 3 bipolar channels EEG data from 9 subjects. Multi-class extensions to FBCSP are also presented to handle the 4-class EEG data in Dataset 2a, namely, Divide-and-Conquer (DC), Pair-Wise (PW), and One-Versus-Rest (OVR) approaches. Two feature selection algorithms are also presented to select discriminative CSP features on Dataset 2b, namely, the Mutual Information-based Best Individual Feature (MIBIF) algorithm, and the Mutual Information-based Rough Set Reduction (MIRSR) algorithm. The single-trial classification accuracies were presented using 10 × 10-fold cross-validations on the training data and session-to-session transfer on the evaluation data from both datasets. Disclosure of the test data labels after the BCI Competition IV showed that the FBCSP algorithm performed relatively the best among the other submitted algorithms and yielded a mean kappa value of 0.569 and 0.600 across all subjects in Datasets 2a and 2b respectively.

A Large Clinical Study on the Ability of Stroke Patients to Use an EEG-Based Motor Imagery Brain-Computer Interface

Brain-computer interface (BCI) technology has the prospects of helping stroke survivors by enabling the interaction with their environment through brain signals rather than through muscles, and restoring motor function by inducing activity-dependent brain plasticity. This paper presents a clinical study on the extent of detectable brain signals from a large population of stroke patients in using EEG-based motor imagery BCI. EEG data were collected from 54 stroke patients whereby finger tapping and motor imagery of the stroke-affected hand were performed by 8 and 46 patients, respectively. EEG data from 11 patients who gave further consent to perform motor imagery were also collected for second calibration and third independent test sessions conducted on separate days. Off-line accuracies of classifying the two classes of EEG from finger tapping or motor imagery of the stroke-affected hand versus the EEG from background rest were then assessed and compared to 16 healthy subjects. The mean off-line accuracy of detecting motor imagery by the 46 patients (μ=0.74) was significantly lower than finger tapping by 8 patients (μ=0.87, p=0.008), but not significantly lower than motor imagery by healthy subjects (μ=0.78, p=0.23). Six stroke patients performed motor imagery at chance level, and no correlation was found between the accuracies of detecting motor imagery and their motor impairment in terms of Fugl-Meyer Assessment (p=0.29). The off-line accuracies of the 11 patients in the second session (μ=0.76) were not significantly different from the first session (μ=0.72, p=0.16), or from the on-line accuracies of the third independent test session (μ=0.82, p=0.14). Hence this study showed that the majority of stroke patients could use EEG-based motor imagery BCI.

Optimizing the Channel Selection and Classification Accuracy in EEG-Based BCI

Multichannel EEG is generally used in brain-computer interfaces (BCIs), whereby performing EEG channel selection 1) improves BCI performance by removing irrelevant or noisy channels and 2) enhances user convenience from the use of lesser channels. This paper proposes a novel sparse common spatial pattern (SCSP) algorithm for EEG channel selection. The proposed SCSP algorithm is formulated as an optimization problem to select the least number of channels within a constraint of classification accuracy. As such, the proposed approach can be customized to yield the best classification accuracy by removing the noisy and irrelevant channels, or retain the least number of channels without compromising the classification accuracy obtained by using all the channels. The proposed SCSP algorithm is evaluated using two motor imagery datasets, one with a moderate number of channels and another with a large number of channels. In both datasets, the proposed SCSP channel selection significantly reduced the number of channels, and outperformed existing channel selection methods based on Fisher criterion, mutual information, support vector machine, common spatial pattern, and regularized common spatial pattern in classification accuracy. The proposed SCSP algorithm also yielded an average improvement of 10% in classification accuracy compared to the use of three channels (C3, C4, and Cz).

A self-training semi-supervised SVM algorithm and its application in an EEG-based brain computer interface speller system

In this paper, we first present a self-training semi-supervised support vector machine (SVM) algorithm and its corresponding model selection method, which are designed to train a classifier with small training data. Next, we prove the convergence of this algorithm. Two examples are presented to demonstrate the validity of our algorithm with model selection. Finally, we apply our algorithm to a data set collected from a P300-based brain computer interface (BCI) speller. This algorithm is shown to be able to significantly reduce training effort of the P300-based BCI speller.

An EEG-Based BCI System for 2-D Cursor Control by Combining Mu/Beta Rhythm and P300 Potential

Two-dimensional cursor control is an important and challenging issue in EEG-based brain-computer interfaces (BCIs). To address this issue, here we propose a new approach by combining two brain signals including Mu/Beta rhythm during motor imagery and P300 potential. In particular, a motor imagery detection mechanism and a P300 potential detection mechanism are devised and integrated such that the user is able to use the two signals to control, respectively, simultaneously, and independently, the horizontal and the vertical movements of the cursor in a specially designed graphic user interface. A real-time BCI system based on this approach is implemented and evaluated through an online experiment involving six subjects performing 2-D control tasks. The results attest to the efficacy of obtaining two independent control signals by the proposed approach. Furthermore, the results show that the system has merit compared with prior systems: it allows cursor movement between arbitrary positions.