Research Items (109)
Objective The use of brain-computer interface in neurofeedback therapy for attention deficit hyperactivity disorder (ADHD) is a relatively new approach. We conducted a randomized controlled trial (RCT) to determine whether an 8-week brain computer interface (BCI)-based attention training program improved inattentive symptoms in children with ADHD compared to a waitlist-control group, and the effects of a subsequent 12-week lower-intensity training. Study design We randomized 172 children aged 6–12 attending an outpatient child psychiatry clinic diagnosed with inattentive or combined subtypes of ADHD and not receiving concurrent pharmacotherapy or behavioral intervention to either the intervention or waitlist-control group. Intervention involved 3 sessions of BCI-based training for 8 weeks, followed by 3 training sessions per month over the subsequent 12 weeks. The waitlist-control group received similar 20-week intervention after a wait-time of 8 weeks. Results The participants’ mean age was 8.6 years (SD = 1.51), with 147 males (85.5%) and 25 females (14.5%). Modified intention to treat analyzes conducted on 163 participants with at least one follow-up rating showed that at 8 weeks, clinician-rated inattentive symptoms on the ADHD-Rating Scale (ADHD-RS) was reduced by 3.5 (SD 3.97) in the intervention group compared to 1.9 (SD 4.42) in the waitlist-control group (between-group difference of 1.6; 95% CI 0.3 to 2.9 p = 0.0177). At the end of the full 20-week treatment, the mean reduction (pre-post BCI) of the pooled group was 3.2 (95% CI 2.4 to 4.1). Conclusion The results suggest that the BCI-based attention training program can improve ADHD symptoms after a minimum of 24 sessions and maintenance training may sustain this improvement. This intervention may be an option for treating milder cases or as an adjunctive treatment.
- Jul 2018
- 2018 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC)
Ballistocardiography (BCG) is a revamped technology for cardiac function monitoring. Detecting individual heart beats in BCG remains a challenging task due to various artifacts and low signal-to-noise ratio, which are not well addressed by conventional approaches based on intuitive observations of BCG waveforms. In this paper, we propose to employ deep learning networks to capture the characteristics of the variations of BCG waveforms within and across individual subjects. Particularly, we design a neural network that combines Convolutional-Neural-Network (CNN) and Extreme Learning Machine (ELM). We test the new learning method on a real BCG data set and show better detection result compared with a state-of-the-art method. We demonstrate how the advanced machine learning technology can learn and detect BCG waveforms robustly.
To detect the mental task of interest, spatial filtering has been widely used to enhance the spatial resolution of electroencephalography (EEG). However, the effectiveness of spatial filtering is undermined due to the significant nonstationarity of EEG. Based on regularization, most of the conventional stationary spatial filter design methods address the nonstationarity at the cost of the interclass discrimination. Moreover, spatial filter optimization is inconsistent with feature extraction when EEG covariance matrices could not be jointly diagonalized due to the regularization. In this paper, we propose a novel framework for a spatial filter design. With Fisher's ratio in feature space directly used as the objective function, the spatial filter optimization is unified with feature extraction. Given its ratio form, the selection of the regularization parameter could be avoided. We evaluate the proposed method on a binary motor imagery data set of 16 subjects, who performed the calibration and test sessions on different days. The experimental results show that the proposed method yields improvement in classification performance for both single broadband and filter bank settings compared with conventional nonunified methods. We also provide a systematic attempt to compare different objective functions in modeling data nonstationarity with simulation studies.
Electrooculogram (EOG) artifact contamination is a common critical issue in general electroencephalogram (EEG) studies as well as in brain computer interface (BCI) research. It is especially challenging when dedicated EOG channels are unavailable or when there are very few EEG channels available for ICA-based ocular artifact removal. It is even more challenging to avoid loss of the signal of interest during the artifact correction process, where the signal of interest can be multiple magnitudes weaker than the artifact. To address these issues, we propose a novel discriminative ocular artifact correction approach for feature learning in EEG analysis.Without extra ocular movement measurements, the artifact is extracted from raw EEG data, which is totally automatic and requires no visual inspection of artifacts. Then, artifact correction is optimized jointly with feature extraction by maximizing oscillatory correlations between trials from the same class and minimizing them between trials from different classes. We evaluate this approach on a real world EEG data set comprising 68 subjects performing cognitive tasks. The results showed that the approach is capable of not only suppressing the artifact components but also improving the discriminative power of a classifier with statistical significance. We also demonstrate that the proposed method addresses the confounding issues induced by ocular movements in cognitive EEG study.
Unobtrusive and long-term monitoring of human vital signs are essential requirements for early diagnosis and prophylaxis due to many reasons, one of the most important being improving the quality of life. Currently, vital signs are continuously monitored through sensors attached to the body, such as multiple electrodes for measuring electrical activity of the heart. Such methods may be undesirable, especially for elderly, infants and other groups of people. In this paper, we introduce an improved technique for measuring heart rate from noisy ballistocardiogram signals acquired from 50 human volunteers in a sitting position using a massage chair. The signals are unobtrusively collected from a microbend fiber optic sensor embedded within the headrest of the chair, and then transmitted to a computer through a Bluetooth connection. The heart rate is computed using the multiresolution analysis of the maximal overlap discrete wavelet transform. The error between the proposed method and the reference ECG is estimated in beats per minute using the mean absolute error, where the system achieved relatively good results (7.31 ± 1.60) despite the large amount of motion artifacts produced owing to the frequent body movements and/or vibrations of the massage chair during stress relief massage. Unlike the complete ensemble empirical mode decomposition algorithm, previously employed for heart rate estimation, the suggested system is much faster. Hence, it can be used in real-time applications.
Opportunistic ambient sensing involves placement of sensors appropriately so that intermittent contact can be made unobtrusively for gathering physiological signals for vital signs. In this paper, we discuss the results of our quality processing system used to extract heart rate from ballistocardiogram signals obtained from a micro-bending fiber optic sensor pressure mat. Visual inspection is used to label data into informative and noninformative classes based on their heart rate information. Five classifiers are employed for the classification process, i.e., random forest, support vector machine, multilayer, feedforward neural network, linear discriminant analysis, and decision tree. To compute the overall effectiveness of quality processing, the informative signals are processed to estimate interbeat intervals. The system was used to process, data collected from 50 human subjects sitting in a massage chair while performing different activities. Opportunistically collected data was obtained from the fiber optic sensor mat placed on the headrest of the massage chair. Using our classification approach, 57.37% of the dataset was able to provide informative signals. On the informative signals, random forest classifier achieves the best classification accuracy with a mean accuracy of 98.99%. The average of the mean absolute error between the estimated heart rate and the reference ECG is reduced from 13.2 to 8.47. Therefore, the proposed system shows a good robustness for opportunistic ambient sensing.
We developed an EEG- and audio-based sleep sensing and enhancing system, called iSleep (interactive Sleep enhancement apparatus). The system adopts a closed-loop approach which optimizes the audio recording selection based on user's sleep status detected through our online EEG computing algorithm. The iSleep prototype comprises two major parts: 1) a sleeping mask integrated with a single channel EEG electrode and amplifier, a pair of stereo earphones and a microcontroller with wireless circuit for control and data streaming; 2) a mobile app to receive EEG signals for online sleep monitoring and audio playback control. In this study we attempt to validate our hypothesis that appropriate audio stimulation in relation to brain state can induce faster onset of sleep and improve the quality of a nap. We conduct experiments on 28 healthy subjects, each undergoing two nap sessions - one with a quiet background and one with our audio-stimulation. We compare the time-to-sleep in both sessions between two groups of subjects, e.g., fast and slow sleep onset groups. The p-value obtained from Wilcoxon Signed Rank Test is 1.22e-04 for slow onset group, which demonstrates that iSleep can significantly reduce the time-to-sleep for people with difficulty in falling sleep.
- Aug 2015
- 2015 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC)
Ballistocardiogram (BCG) is a vital sign of ballistic forces generated by each heartbeat. With the advancements in related sensor and computing technologies in recent years, BCG has become far more accessible and thus regained its interest in both research and industry fields. Here we would like to promote the system modelling approach to BCG computing that allows to explore the underlying association between BCG and other physiological signals such as electrocardiogram (ECG). This is in contrast to most of the existing works in the related signal processing domain, which focus on detecting heart rate only. The system modelling approach may eventually improve the clinical significance of the BCG by extracting deeply embedded information. Towards this goal, here we present our preliminary study where we design a Wavelet-based temporal-frequency system model for associating BCG and ECG. To validate the model, we also collect simultaneous BCG and ECG recordings from 4 healthy subjects. We use the system model to build a BCG to ECG predicting algorithm. We demonstrate that this temporal-frequency model and algorithm is far superior, in terms of accuracy, to the naïve method of linear modelling.
- Aug 2015
- 2015 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC)
Non-intrusiveness is one of the advantages of in-bed optical sensor device for monitoring vital signs, including heart rate and respiratory rate. Estimating respiratory rate reliably using such sensors, however, is challenging, due to body movement, signal variation according to different subjects or body positions, etc. This paper presents a method for reliable respiratory rate estimation for FBG optical sensors by introducing signal quality estimation. The method estimates the quality of the signal waveform by detecting regularly repetitive patterns using proposed spectrum and cepstrum analysis. Multiple window sizes are used to cater for a wide range of target respiratory rates. Furthermore, the readings of multiple sensors are fused to derive a final respiratory rate. Experiments with 12 subjects and 2 body positions were conducted using polysomnography belt signal as groundtruth. The results demonstrated the effectiveness of the method.
This paper presents a method of extracting Pulse Transit Time (PTT) from multimodal pulsatile signals, such as Electrocardiography (ECG) and Photoplethysmograph (PPG), to facilitate research on estimating Blood Pressure based on PTT. We address the robustness issues of PTT extraction, since the pulsatile signals are subject to distortions due to sensor errors or subject movements. A cepstrum analysis method is proposed to estimate the pulse rate from each modality, and then a signal quality assessment is conducted by verifying the conformance of the pulse rate with the pulse time for a single modality as well as across multiple modalities. Erroneous peaks in the distorted pulsatile signal can be rejected by the quality screening step. Experiments were conducted using PhysioNet database MIMIC-II comprising ECG, PPG and Arterial Blood Pressure (ABP) signals, and the results demonstrated the effectiveness of the proposed method and furthermore provided insights to the MIMIC-II database.
Background: There is growing evidence that cognitive training (CT) can improve the cognitive functioning of the elderly. CT may be influenced by cultural and linguistic factors, but research examining CT programs has mostly been conducted on Western populations. We have developed an innovative electroencephalography (EEG)-based brain-computer interface (BCI) CT program that has shown preliminary efficacy in improving cognition in 32 healthy English-speaking elderly adults in Singapore. In this second pilot trial, we examine the acceptability, safety, and preliminary efficacy of our BCI CT program in healthy Chinese-speaking Singaporean elderly. Methods: Thirty-nine elderly participants were randomized into intervention (n=21) and wait-list control (n=18) arms. Intervention consisted of 24 half-hour sessions with our BCI-based CT training system to be completed in 8 weeks; the control arm received the same intervention after an initial 8-week waiting period. At the end of the training, a usability and acceptability questionnaire was administered. Efficacy was measured using the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS), which was translated and culturally adapted for the Chinese-speaking local population. Users were asked about any adverse events experienced after each session as a safety measure. Results: The training was deemed easily usable and acceptable by senior users. The median difference in the change scores pre- and post-training of the modified RBANS total score was 8.0 (95% confidence interval [CI]: 0.0-16.0, P=0.042) higher in the intervention arm than waitlist control, while the mean difference was 9.0 (95% CI: 1.7-16.2, P=0.017). Ten (30.3%) participants reported a total of 16 adverse events - all of which were graded "mild" except for one graded "moderate". Conclusion: Our BCI training system shows potential in improving cognition in both English- and Chinese-speaking elderly, and deserves further evaluation in a Phase III trial. Overall, participants responded positively on the usability and acceptability questionnaire.
There is a method and a system for concentration detection. The method for concentration detection includes the steps of extracting temporal features from brain signals; classifying the extracted temporal features using a classifier to give a score x1; extracting spectral-spatial features from brain signals; selecting spectral-spatial features containing discriminative information between concentration and non-concentration states from the set of extracted spectral-spatial features; classifying the selected spectral-spatial features using a classifier to give a score x2; combining the scores x1 and x2 to give a single score; and determining if the subject is in a concentration state based on the single score.
A method or system for classifying brain signals in a BCI. The system comprises a model building unit for building a subject-independent model using labelled brain signals from a pool of subjects.
Objective: Session-to-session nonstationarity is inherent in brain-computer interfaces based on electroencephalography. The objective of this paper is to quantify the mismatch between the training model and test data caused by nonstationarity and to adapt the model towards minimizing the mismatch. Approach: We employ a tensor model to estimate the mismatch in a semi-supervised manner, and the estimate is regularized in the discriminative objective function. Main results: The performance of the proposed adaptation method was evaluated on a dataset recorded from 16 subjects performing motor imagery tasks on different days. The classification results validated the advantage of the proposed method in comparison with other regularization-based or spatial filter adaptation approaches. Experimental results also showed that there is a significant correlation between the quantified mismatch and the classification accuracy. Significance: The proposed method approached the nonstationarity issue from the perspective of data-model mismatch, which is more direct than data variation measurement. The results also demonstrated that the proposed method is effective in enhancing the performance of the feature extraction model.
To address the nonstationarity issue in EEG-based brain computer interface (BCI), the computational model trained using the training data needs to adapt to the data from the test sessions. In this paper, we propose a novel adaptation approach based on the divergence framework. Cross-session changes can be taken into consideration by searching the discriminative subspaces for test data on the manifold of orthogonal matrices in a semi-supervised manner. Subsequently, the feature space becomes more consistent across sessions and classifiers performance can be enhanced. Experimental results show that the proposed adaptation method yields improvements in classification performance.
Physiological sensor based workload estimation technology provides a real-time means for assessing cognitive workload and has a broad range of applications in cognitive ergonomics, mental health monitoring, etc. In this paper we report a study on detecting changes in workload using multi-modality physiological sensors and a novel feature extrac-tion and classification algorithm. We conducted a cognitive workload experiment involving multiple subjects and collected an extensive data set of EEG, ECG and GSR signals. We show that the GSR signal is consistent with the variations of cognitive workload in 75% of the samples. To explore cardiac patterns in ECG that are potentially correlated with the cognitive workload process, we computed various heart-rate-variability features. To extract neuronal activity patterns in EEG related to cognitive workload, we introduced a filter bank common spatial pattern filtering technique. As there can be large variations in e.g. individual responses to the cognitive workload, we propose a large margin unbiased recursive feature extraction and regression method. Our leave-one-subject-out cross validation test shows that, using the proposed method, EEG can provide significantly better prediction of the cognitive workload variation than ECG, with 87.5% vs 62.5% in accuracy rate.
This paper presents a method of estimating heart rate from arrays of fiber Bragg grating (FBG) sensors embedded in a mat. A cepstral domain signal analysis technique is proposed to characterize Ballistocardiogram (BCG) signals. With this technique, the average heart beat intervals can be estimated by detecting the dominant peaks in the cepstrum, and the signals of multiple sensors can be fused together to obtain higher signal to noise ratio than each individual sensor. Experiments were conducted with 10 human subjects lying on 2 different postures on a bed. The estimated heart rate from BCG was compared with heart rate ground truth from ECG, and the mean error of estimation obtained is below 1 beat per minute (BPM). The results show that the proposed fusion method can achieve promising heart rate measurement accuracy and robustness against various sensor contact conditions.
Sleep has been shown to be imperative for the health and well-being of an individual. To design intelligent sleep management tools, such as the music-induce sleep-aid device, automatic detection of sleep onset is critical. In this work, we propose a simple yet accurate method for sleep onset prediction, which merely relies on Electroencephalogram (EEG) signal acquired from a single frontal electrode in a wireless headband. The proposed method first extracts energy power ratio of theta (4-8Hz) and alpha (8-12Hz) bands along a 3-second shifting window, then calculates the slow wave of each frequency band along the time domain. The resulting slow waves are then fed to a rule-based engine for sleep onset detection. To evaluate the effectiveness of the approach, polysomnographic (PSG) and headband EEG signals were obtained from 20 healthy adults, each of which underwent 2 sessions of sleep events. In total, data from 40 sleep events were collected. Each recording was then analyzed offline by a PSG technologist via visual observation of PSG waveforms, who annotated sleep stages N1 and N2 by using the American Academy of Sleep Medicine (AASM) scoring rules. Using this as the gold standard, our approach achieved a 87.5% accuracy for sleep onset detection. The result is better or at least comparable to the other state of the art methods which use either multi-or single-channel based data. The approach has laid down the foundations for our future work on developing intelligent sleep aid devices.
- Aug 2014
We report results from a clinical trial for monitoring respiration and cardiac activity of patients during sleep using microbend fiber sensor. This sensor is used to acquire respiratory and heart beat information. We have collected reference data from standard Polysomnography and data from microbend fiber sensor on 22 patients. A new algorithm is developed to calculate breathing rate and heart rate simultaneously. The Bland-Altman analysis demonstrates the measurements have good accuracy for monitoring purposes compared with the standard Polysomnography. An accuracy of 1.06bpm for breathing rate and 3.32bpm for heart rate has been validated for 30s averaging time although there were significant signal distortions under sleep conditions.
The non-stationarity inherent across sessions recorded on different days poses a major challenge for practical electroencephalography (EEG)-based Brain Computer Interface (BCI) systems. To address this issue, the computational model trained using the training data needs to adapt to the data from the test sessions. In this paper, we propose a novel approach to compute the variations between labelled training data and a batch of unlabelled test data based on the geodesic-distance of the discriminative subspaces of EEG data on the Grassmann manifold. Subsequently, spatial filters can be updated and features that are invariant against such variations can be obtained using a subset of training data that is closer to the test data. Experimental results show that the proposed adaptation method yielded improvements in classification performance.
We have made use of the microbending fiber optic sensor to capture ballistocardiographic signals or data for vital signs monitoring in ambient settings, with applications ranging from serious games to ambient assistive living for ageing at home for the elderly. To remove noise and extract the vital signs, the first step of the signal data processing is filtering the signal. In this paper we consider the properties of the digital filter for filtering ballistocardiographic signals. The vital signs waveforms are derived from raw data captured by optical transducers that are placed in ambient locations that are in contact with, but not worn by the subject. Data has been collected from various locations and positions and a detailed trial has been conducted for one of these positions. We iteratively improve the filter design so as to lead to the best parameters. The baseline filter performed reasonably well on data collected in a trial study, with a mean error rate less than 10% for half of the subjects and below 20% for three quarters of the subjects. We also present results of an improved filter that improves the performance both in terms of responsiveness and sensitivity. The improved filter demonstrates consistently less than 12% mean error rate. Principles gleaned from this study may also be applied in designing filters for other types of sensors and for other applications in healthcare.
Unlabelled: Cognitive decline in aging is a pressing issue associated with significant healthcare costs and deterioration in quality of life. Previously, we reported the successful use of a novel brain-computer interface (BCI) training system in improving symptoms of attention deficit hyperactivity disorder. Here, we examine the feasibility of the BCI system with a new game that incorporates memory training in improving memory and attention in a pilot sample of healthy elderly. This study investigates the safety, usability and acceptability of our BCI system to elderly, and obtains an efficacy estimate to warrant a phase III trial. Thirty-one healthy elderly were randomized into intervention (n = 15) and waitlist control arms (n = 16). Intervention consisted of an 8-week training comprising 24 half-hour sessions. A usability and acceptability questionnaire was administered at the end of training. Safety was investigated by querying users about adverse events after every session. Efficacy of the system was measured by the change of total score from the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) before and after training. Feedback on the usability and acceptability questionnaire was positive. No adverse events were reported for all participants across all sessions. Though the median difference in the RBANS change scores between arms was not statistically significant, an effect size of 0.6SD was obtained, which reflects potential clinical utility according to Simon's randomized phase II trial design. Pooled data from both arms also showed that the median change in total scores pre and post-training was statistically significant (Mdn = 4.0; p<0.001). Specifically, there were significant improvements in immediate memory (p = 0.038), visuospatial/constructional (p = 0.014), attention (p = 0.039), and delayed memory (p<0.001) scores. Our BCI-based system shows promise in improving memory and attention in healthy elderly, and appears to be safe, user-friendly and acceptable to senior users. Given the efficacy signal, a phase III trial is warranted. Trial registration: ClinicalTrials.gov NCT01661894.
The performance of Brain-Computer Interface (BCI) applications are sometimes hindered by non-stationarity in the EEG data from sessions on different days. This paper proposes an algorithm for adaptive training of a SVM classifier to address the non-stationarity in EEG by adapting the kernel to data from subsequent sessions. The kernel width parameter of the kernel function of the SVM classifier is adapted using an information theoretic cost function based on minimum error entropy (MEE). An experiment is performed using the proposed method on EEG data collected without feedback from 12 healthy subjects in two sessions on separate days. The results using the proposed method yielded a mean accuracy of 75%, which is significantly better compared to the baseline result of 67% without kernel adaptation (P=0.00029).
Effective learning and recovery of relevant source brain activity patterns is a major challenge to brain-computer interface using scalp EEG. Various spatial filtering solutions have been developed. Most current methods estimate an instantaneous demixing with the assumption of uncorrelatedness of the source signals. However, recent evidence in neuroscience suggests that multiple brain regions cooperate, especially during motor imagery, a major modality of brain activity for brain-computer interface. In this sense, methods that assume uncorrelatedness of the sources become inaccurate. Therefore, we are promoting a new methodology that considers both volume conduction effect and signal propagation between multiple brain regions. Specifically, we propose a novel discriminative algorithm for joint learning of propagation and spatial pattern with an iterative optimization solution. To validate the new methodology, we conduct experiments involving 16 healthy subjects and perform numerical analysis of the proposed algorithm for EEG classification in motor imagery brain-computer interface. Results from extensive analysis validate the effectiveness of the new methodology with high statistical significance.
- Jul 2013
Ballistocardiography (BCG) is a promising unobtrusive method for home e-healthcare systems, and has attracted increasing interest in recent years along with technological advances in related biomedical, electrical engineering and computer science fields. While existing systems have investigated the efficacy of BCG setups in bed, backrest, seat or scale positions, we propose to study BCG in headrest position that will allow new practical and portable applications. To this end, we designed and implemented a multi-modality sensing system including a high-sensitivity microbend fiber optic BCG sensor. In this preliminary study, we have collected multi-modality physiological data on 3 human subjects. We ran extensive analysis on BCG in correlation with ECG, and identified special characteristics of the signal in the new BCG setup. The result suggests that new appropriate computing techniques are necessary for accurately recovering the heart beat signal. Therefore, we developed a novel algorithm for heart beat detection. We evaluate the algorithm with the data and demonstrate that it can accurately compute heart rate intervals in the headrest BCG despite significant signal distortion.
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.
A system and method for processing brain signals in a BCI system. The method of processing brain signals in a BCI system includes the steps of processing the brain signals for control state detection to determine if a subject intends to use the BCI system; and processing the brain signals for command recognition if the control state detection method determines that the subject intends to use the BCI system.
Objective: The non-stationary nature of EEG poses a major challenge to robust operation of brain-computer interfaces (BCIs). The objective of this paper is to propose and investigate a computational method to address non-stationarity in EEG classification. Approach: We developed a novel dynamically weighted ensemble classification (DWEC) framework whereby an ensemble of multiple classifiers are trained on clustered features. The decisions from these multiple classifiers are dynamically combined based on the distances of the cluster centres to each test data sample being classified. Main results: The clusters of the feature space from the second session spanned a different space compared to the clusters of the feature space from the first session which highlights the processes of session-to-session non-stationarity. The session-to-session performance of the proposed DWEC method was evaluated on two datasets. The results on publicly available BCI Competition IV dataset 2A yielded a significantly higher mean accuracy of 81.48% compared to 75.9% from the baseline support vector machine (SVM) classifier without dynamic weighting. Results on the data collected from our twelve in-house subjects yielded a significantly higher mean accuracy of 73% compared to 69.4% from the baseline SVM classifier without dynamic weighting. Significance: The cluster based analysis provides insight into session-to-session non-stationarity in EEG data. The results demonstrate the effectiveness of the proposed method in addressing non-stationarity in EEG data for the operation of a BCI.
Motion intention can be detected from human Electroencephalography (EEG) signals through BCI, which can facilitate motor motion control for disabled or paralyzed people. However, the continuous use of BCI is hindered by the non-stationarity of the EEG signals. This paper proposes a method to identify the EEG signal components that can be used to train a classifier to address the non-stationarity issue. The proposed method is based on Transfer Component Analysis (TCA). TCA seeks to locate components that can be transferred across domains in a Reproducing Kernel Hilbert Space (RKHS). The distributions associated with data are closer to each other in the subspaces spanned by the identified transfer components. Therefore, typical machine learning techniques can be applied in the subspace spanned by these transfer components. This results in classifiers that can be trained on the source domain and tested on the target domain. The proposed Stationary Transfer Component Analysis (STCA) method is compared with Stationary Sub-space Analysis (SSA) on the BCI competition IV dataset 2a. The results show significant improvements over the baseline case and the results are better than those produced by SSA.
Attention deficit hyperactivity disorder (ADHD) symptoms can be difficult to treat. We previously reported that a 20-session brain-computer interface (BCI) attention training programme improved ADHD symptoms. Here, we investigated a new more intensive BCI-based attention training game system on 20 unmedicated ADHD children (16 males, 4 females) with significant inattentive symptoms (combined and inattentive ADHD subtypes). This new system monitored attention through a head band with dry EEG sensors, which was used to drive a feed forward game. The system was calibrated for each user by measuring the EEG parameters during a Stroop task. Treatment consisted of an 8-week training comprising 24 sessions followed by 3 once-monthly booster training sessions. Following intervention, both parent-rated inattentive and hyperactive-impulsive symptoms on the ADHD Rating Scale showed significant improvement. At week 8, the mean improvement was −4.6 (5.9) and −4.7 (5.6) respectively for inattentive symptoms and hyperactive-impulsive symptoms (both p<0.01). Cohen’s d effect size for inattentive symptoms was large at 0.78 at week 8 and 0.84 at week 24 (post-boosters). Further analysis showed that the change in the EEG based BCI ADHD severity measure correlated with the change ADHD Rating Scale scores. The BCI-based attention training game system is a potential new treatment for ADHD. Trial Registration ClinicalTrials.gov NCT01344044
The common spatial pattern (CSP) algorithm is effective in decoding the spatial patterns of the corresponding neuronal activities from electroencephalogram (EEG) signal patterns in brain–computer interfaces (BCIs). However, its effectiveness depends on the subject-specific time segment relative to the visual cue and on the temporal frequency band that is often selected manually or heuristically. This paper presents a novel statistical method to automatically select the optimal subject-specific time segment and temporal frequency band based on the mutual information between the spatial–temporal patterns from the EEG signals and the corresponding neuronal activities. The proposed method comprises four progressive stages: multi-time segment and temporal frequency band-pass filtering, CSP spatial filtering, mutual information-based feature selection and naïve Bayesian classification. The proposed mutual information-based selection of optimal spatial–temporal patterns and its one-versus-rest multi-class extension were evaluated on single-trial EEG from the BCI Competition IV Datasets IIb and IIa respectively. The results showed that the proposed method yielded relatively better session-to-session classification results compared against the best submission.
Feature extraction has been a crucial and challenging task for EEG-based BCI applications mainly due to the problems of high-dimensionality and high noise level of EEG signals. In this paper we developed a novel feature extraction algorithm for EEG-based emotion detection problem. The proposed algorithm is derived from viewing EEG signals as the activation/deactivation of sources specific to the brain activities of interest. For binary classification problem, to be more specific, we consider the EEG signals for the two types of brain activities as characterized by the activation/deactivation of two discriminatory sources in the brain, with one source activated and the other one deactivated for one particular type of brain activities. The proposed algorithm, termed Asymmetric Spatial Pattern (ASP), extracts pairs of spatial filters, with each filter corresponding to only one of the two sources. The idea of ASP is neurologically plausible for certain situations. For example, according to the valence hypothesis of emotion, the left hemisphere is more activated in positive emotions and the right hemisphere is more activated in negative emotions. The effectiveness of the proposed algorithm is confirmed by application to real data for two types of EEG-based emotion detection problems: arousal detection (strong v.s. calm), and valence detection (positive v.s. negative). Experimental results on the real data also show that some of the asymmetric spatial patterns by ASP are consistent with the current neurophysiological findings on brain emotion processing.
This paper addresses an important problem known as EEG non-stationarity in Brain-computer Interfacing. We propose a novel technique called Dynamically Weighted Classification with Clustering (DWCC), which explores hidden states in non-stationary EEG using a modified k-means clustering method by combining cosine distance measure and mutual information criterion. DWCC builds a set of classifiers, one for each pair of clusters from different classes. A dynamically-weighted classifier ensemble network is trained to combine the outputs of the classifiers, where we propose to dynamically assign the weight of a classifier for each test sample based on its distances to the cluster centres associated with the classifier. Experimental results on publicly available BCI Competition IV Dataset 2a yielded a mean accuracy of 81.5% which is statistically significant (t-test p
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.
This paper investigates how to apply active learning for the classification of motor imagery electroencephalography (EEG) signals to boost the performance for small training size. A new criterion is proposed to select the most representative and informative queries. The candidates are firstly chosen from the samples close to the center of the cluster that has the highest impurity of classes. A predefined number of such candidates and classifiers are forwardly buffered. Subsequently, the query is chosen such that the buffered classifiers can backward maximize the classification errors on labeled data. Experimental results conducted on the BCI competition IV data set IVb show the superior performance of the proposed active learning scheme, which is on average 5.12% higher in accuracy than that of the passive method by choosing the training size from 28 to 112.
The injection of emotional intelligence in human-computer interfaces is necessary for computer applications to appear intelligent when interacting with people. With the recent development of brain imaging techniques and brain-computer interfaces, computers can actually take a look inside users' head to observe their emotional states. This paper presents an EEG-based emotion detection system which detects emotional states based on short EEG segments of 1s. A novel feature extraction algorithm termed asymmetric spatial filtering is proposed to extract features from high dimensional EEG data. The effectiveness of the proposed method is tested for two types of emotion detection problems on data from five subjects.
Detecting motor imagery activities versus non-control in brain signals is the basis of self-paced brain-computer interfaces (BCIs), but also poses a considerable challenge to signal processing due to the complex and non-stationary characteristics of motor imagery as well as non-control. This paper presents a self-paced BCI based on a robust learning mechanism that extracts and selects spatio-spectral features for differentiating multiple EEG classes. It also employs a non-linear regression and post-processing technique for predicting the time-series of class labels from the spatio-spectral features. The method was validated in the BCI Competition IV on Dataset I where it produced the lowest prediction error of class labels continuously. This report also presents and discusses analysis of the method using the competition data set.
This paper proposes a novel active learning method for the classification of motor imagery electroencephalogram (EEG) signals. Specifically, we propose an iterative clustering and support vector-based criterion to select samples of high-confidence to construct a robust training set. The common spatial pattern (CSP)-based features are iteratively clustered till the number of support vectors in the cluster is less than a predefined threshold. A predefined number of samples close to the cluster centers are chosen. When such clusters cannot be found, the samples that are of farthest distances to a group of support vectors of class “0” and “1” are alternately chosen. Experimental results on BCI competition IV dataset IIb show superior performance compared with a baseline method, which is 9% increase in accuracy averaged across subjects and training sizes.
Brain-computer interface (BCI) technology has the prospects of helping stroke survivors by enabling the interaction with their environ ment 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 (mu=0.74) was significantly lower than finger tapping by 8 patients (mu=0.87, p=0.008), but not significantly lower than motor imagery by healthy subjects (mu=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 (mu=0.76) were not significantly different from the first session (mu=0.72, p=0.16), or from the on-line accuracies of the third independent test session (mu=0.82, p=0.14). Hence this study showed that the majority of stroke patients could use EEG-based motor imagery BCI.
Inherent changes that appear in brain signals when transferring from calibration to feedback sessions are a challenging but critical issue in brain-computer interface (BCI) applications. While previous studies have mostly focused on the adaptation of classifiers, in this paper we study the feasibility and the importance of the adaptation of feature extraction in a self-paced BCI paradigm. First, we conduct calibration and feedback training on able-bodied naïve subjects using a new self-paced motor imagery BCI including the idle state. The online results suggest that the feature space constructed from calibration data may become ineffective during feedback sessions. Hence, we propose a new supervised method that learns from a feedback session to construct a more appropriate feature space, on the basis of the maximum mutual information principle between feedback signal, target signal and EEG. Specifically, we formulate the learning objective as maximizing a kernel-based mutual information estimate with respect to the spatial-spectral filtering parameters. We then derive a gradient-based optimization algorithm for the learning task. An experimental study is conducted using offline simulation. The results show that the proposed method is able to construct effective feature spaces to capture the discriminative information in feedback training data and, consequently, the prediction error can be significantly reduced using the new features.
The Filter Bank Common Spatial Pattern (FBCSP) algorithm employs multiple spatial filters to automatically select key temporal-spatial discriminative EEG characteristics and the Naïve Bayesian Parzen Window (NBPW) classifier using offline learning in EEG-based Brain-Computer Interfaces (BCI). However, it has yet to address the non-stationarity inherent in the EEG between the initial calibration session and subsequent online sessions. This paper presents the FBCSP that employs the NBPW classifier using online adaptive learning that augments the training data with available labeled data during online sessions. However, employing semi-supervised learning that simply augments the training data with available data using predicted labels can be detrimental to the classification accuracy. Hence, this paper presents the FBCSP using online semi-supervised learning that augments the training data with available data that matches the probabilistic model captured by the NBPW classifier using predicted labels. The performances of FBCSP using online adaptive and semi-supervised learning are evaluated on the BCI Competition IV datasets IIa and IIb and compared to the FBCSP using offline learning. The results showed that the FBCSP using online semi-supervised learning yielded relatively better session-to-session classification results compared against the FBCSP using offline learning. The FBCSP using online adaptive learning on true labels yielded the best results in both datasets, but the FBCSP using online semi-supervised learning on predicted labels is more practical in BCI applications where the true labels are not available.
The Filter Bank Common Spatial Pattern (FBCSP) algorithm constructs and selects subject-specific discriminative CSP features from a filter bank of spatial- temporal filters in a motor imagery brain-computer interface (MI-BCI). However, information from other types of features could be extracted and combined with CSP features to enhance the classification performance. Hence this paper proposes a Filter Bank Feature Combination (FBFC) approach and investigates the use of CSP and Phase Lock Value (PLV) features, where the latter measures the phase synchronization between the EEG electrodes. The performance of the FBFC using CSP and PLV features is evaluated on four-class motor imageries from the publicly available BCI Competition IV Dataset IIa. The experimental results showed that the proposed FBFC using CSP and PLV features yielded a significant improvement in cross-validation accuracies on the training data (p=0.008) and better session-to-session transfer accuracies to the evaluation data compared to the use of CSP features using the FBCSP algorithm. This motivates the research of FBFC using a battery of other features that could possibly benefit EEG-based BCIs and multi-modal BCI systems.
- Jun 2011
This paper presents a novel approach which uses brain computer interface (BCI) technology to translate the user's mental conditions, especially the attention state, into game control. Leveraging on BCI engine to measure a user's attention level to control a virtual hand's movement and utilizing D animation technique, the proposed system is significant for training those who suffering from Attention Deficit Hyperactivity Disorder (ADHD). Comparing to robotic based system, the proposed system is cost-effective, interesting, and ease of use. It also can be extended for rehabilitating the people with neurological disorders, such as those debilitating traumatic events. Potentially, millions people may benefit from the system. The system structure and experimental results will be illustrated in this paper. To our knowledge, no same system is reported yet.
The Filter Bank Common Spatial Pattern (FBCSP) algorithm employs multiple spatial filters across a bank of band-pass filtered EEC using the CSP algorithm. This is as opposed to the commonly used single spatial filter from band-pass filtered EEC. Hence, the FBCSP yields improved performance in autonomous selection of key temporal-spatial discriminative EEC characteristics in motor imagery-based Brain-Computer Interfaces (MI-BCI). However, the multiple spatial filtering involves multiple estimations of covariance matrices across the different frequency bands. Thus, the use of multiple spatial filters increases the sensitivity of the FBCSP algorithm to noise, artifacts and outliers compared to the CSP algorithm. Furthermore, the multiple spatial patterns are also less interpretable than a single spatial pattern. Hence this paper proposes a Composite FBCSP algorithm that employs a single spatial filter instead of multiple spatial filters. The composite spatial filter is computed from a weighted sum of covariance matrices whereby the weights are determined from the mutual information across selected frequency band. The performance of the Composite FBCSP is compared to the FBCSP on a publicly available dataset and data collected from 5 healthy subjects using session-to-session transfer kappa values on the independent test data. The results revealed improvements in accuracy and interpretability in the spatial patterns.
We present a new linear discriminant analysis method based on information theory, where the mutual information between linearly transformed input data and the class labels is maximized. First, we introduce a kernel-based estimate of mutual information with a variable kernel size. Furthermore, we devise a learning algorithm that maximizes the mutual information w.r.t. the linear transformation. Two experiments are conducted: the first one uses a toy problem to visualize and compare the transformation vectors in the original input space; the second one evaluates the performance of the method for classification by employing cross-validation tests on four datasets from the UCI repository. Various classifiers are investigated. Our results show that this method can significantly boost class separability over conventional methods, especially for nonlinear classification.
This paper proposes a feature extraction method for motor imagery brain-computer interface (BCI) using electroencephalogram. We consider the primary neurophysiologic phenomenon of motor imagery, termed event-related desynchronization, and formulate the learning task for feature extraction as maximizing the mutual information between the spatio-spectral filtering parameters and the class labels. After introducing a nonparametric estimate of mutual information, a gradient-based learning algorithm is devised to efficiently optimize the spatial filters in conjunction with a band-pass filter. The proposed method is compared with two existing methods on real data: a BCI Competition IV dataset as well as our data collected from seven human subjects. The results indicate the superior performance of the method for motor imagery classification, as it produced higher classification accuracy with statistical significance (≥95% confidence level) in most cases.
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.
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.
This paper addresses an important issue in a self-paced brain-computer interface (BCI): constructing subject-specific continuous control signal. To this end, we propose an alternative to the conventional regression/classification-based mechanism for building the transformation from EEG features into a univariate control signal. Based on information theory, the mechanism formulates the optimum transformation as maximizing the mutual information between the control signal and the mental state. We introduce a non-parametric mutual information estimate for general output distribution, and then develop a gradient-based algorithm to optimize the transformation using training data. We conduct an offline simulation study using motor imagery data from the BCI Competition IV Data Set I. The results show that the learning algorithm converged quickly, and the proposed method yielded significantly higher BCI performance than the conventional mechanism.
We propose a novel linear discriminant analysis method and demonstrate its superiority over existing linear methods. Based on information theory, we introduce a non-parametric estimate of mutual information with variable kernel bandwidth. Furthermore, we derive a gradient-based optimization algorithm for learning the optimal linear reduction vectors which maximizes the mutual information estimate. We evaluate the proposed method by running cross-validation on 2 data sets from the UCI repository, together with linear and nonlinear SVMs as classifiers. The result attests to the superority of the method over conventional LDA and its variant, aPAC.
This clinical study investigates the ability of hemiparetic stroke patients in operating EEG-based motor imagery brain-computer interface (MI-BCI). It also assesses the efficacy in motor improvements on the stroke-affected upper limb using EEG-based MI-BCI with robotic feedback neurorehabilitation compared to robotic rehabilitation that delivers movement therapy. 54 hemiparetic stroke patients with mean age of 51.8 and baseline Fugl-Meyer Assessment (FMA) 14.9 (out of 66, higher = better) were recruited. Results showed that 48 subjects (89%) operated EEG-based MI-BCI better than at chance level, and their ability to operate EEG-based MI-BCI is not correlated to their baseline FMA (r=0.358). Those subjects who gave consent are randomly assigned to each group (N=11 and 14) for 12 1-hour rehabilitation sessions for 4 weeks. Significant gains in FMA scores were observed in both groups at post-rehabilitation (4.5, 6.2; p=0.032, 0.003) and 2-month post-rehabilitation (5.3, 7.3; p=0.020, 0.013), but no significant differences were observed between groups (p=0.512, 0.550). Hence, this study showed evidences that a majority of hemiparetic stroke patients can operate EEG-based MI-BCI, and that EEG-based MI-BCI with robotic feedback neurorehabilitation is effective in restoring upper extremities motor function in stroke.
In this paper, a hybrid EEG-based brain computer interface (BCI) is designed for two-dimensional cursor control. In our approach, two brain activity patterns, i.e., motor imagery and P300 potential, are used for controlling the horizontal and the vertical movements of the cursor respectively. A real-time BCI system based on this approach is implemented and evaluated through an online experiment. Six subjects attending this experiment can perform 2-D cursor control effectively. Our experimental results show that the system has the following merits compared with prior systems: 1) it does not rely on intensive user training; 2) it allows cursor movement between arbitrary positions.
Optimum linear transformation under mixture of zero-mean Gaussian conditions is an intriguing problem, especially in learning discriminative spatial components in motor imagery EEG for building brain computer interfaces. However, it is not well addressed in the past. In this paper, we study optimum linear transformation under mixture of zero-mean Gaussian. In particular, we formulate optimum transformation as a Bhattacharyya error bound minimization problem, and derive a numerical solution to estimate the bound from training samples. Based on the solution, we develop an algorithm for selecting optimum linear transformation. The proposed method is evaluated, in comparison with the state-of-the-art methods, using a publicly available data set of motor imagery EEG. The results attest to the superiority of the method for detecting motor imagery.
Majority of children with attention deficit hyperactivity disorder (ADHD) have significant inattentive symptoms. We developed a progressive series of activities involving brain-computer interface-based games which could train users to improve their concentration. This pilot study investigated if the intervention could be utilized in children and if it could improve inattentive symptoms of ADHD. Ten medication-naive children aged 7 to 12 diagnosed with ADHD (combined or inattentive subtypes) received 20 sessions of therapy over a 10-week period. They were compared with age- and gendermatched controls. Both parent and teacher-rated inattentive score on the ADHD Rating Scale-IV improved more in the intervention group. A larger scale trial is warranted to further investigate the efficacy of our treatment programme in treating ADHD.
This paper investigates the classification of multi-class motor imagery for electroencephalogram (EEG)-based Brain-Computer Interface (BCI) using the Filter Bank Common Spatial Pattern (FBCSP) algorithm. The FBCSP algorithm classifies EEG measurements from features constructed using subject-specific temporal-spatial filters. However, the FBCSP algorithm is limited to binary-class motor imagery. Hence, this paper proposes 3 approaches of multi-class extension to the FBCSP algorithm: One-versus-Rest, Pair-Wise and Divide-and-Conquer. These approaches decompose the multi-class problem into several binary-class problems. The study is conducted on the BCI Competition IV dataset IIa, which comprises single-trial EEG data from 9 subjects performing 4-class motor imagery of left-hand, right-hand, foot and tongue actions. The results showed that the multi-class FBCSP algorithm could extract features that matched neurophysiological knowledge, and yielded the best performance on the evaluation data compared to other international submissions.
The Filter Bank Common Spatial Pattern (FBCSP) algorithm performs autonomous selection of key temporal-spatial discriminative EEG characteristics in motor imagery-based Brain Computer Interfaces (MI-BCI). However, FBCSP is sensitive to outliers because it involves multiple estimations of covariance matrices from EEG measurements. This paper proposes a Robust FBCSP (RFBCSP) algorithm whereby the estimates of the covariance matrices are replaced with the robust Minimum Covariance Determinant (MCD) estimator. The performance of RFBCSP is investigated on a publicly available dataset and compared against FBCSP using 10x10-fold cross-validation accuracies on training data, and session-to-session transfer kappa values on independent test data. The results showed that RFBCSP yielded improvements in certain subjects and slight improvement in overall performance across subjects. Analysis on one subject who improved suggested that outliers were excluded from the robust covariance matrices estimation. These results revealed a promising direction of RFBCSP for robust classifications of EEG measurements in MI-BCI.
This paper addresses the issue of selecting optimal spatio-spectral features, which is key to high performance motor imagery (MI) classification that is in turn one of the central topics in EEG-based brain computer interfaces. In particular, this work proposes a novel method which first formulates the selection of features as maximizing mutual information between class labels and features. It then uses a robust estimate of mutual information, within a filter-bank and common spatial pattern feature extraction framework, to select an effective feature set. We have assessed the proposed method on both BCI Competition IV Set I and a separate data set collected in our lab from 7 healthy subjects. The results indicate the method is effective in selecting optimal spatial-spectral features for classification.
Non-invasive EEG-based motor imagery brain-computer interface (MI-BCI) holds promise to effectively restore motor control to stroke survivors. This clinical study investigates the effects of MI-BCI for upper limb robotic rehabilitation compared to standard robotic rehabilitation. The subjects are hemiparetic stroke patients with mean age of 50.2 and baseline Fugl-Meyer (FM) score 29.7 (out of 66, higher = better) randomly assigned to each group respectively (N = 8 and 10). Each subject underwent 12 sessions of 1-hour rehabilitation for 4 weeks. Significant gains in FM scores were observed in both groups at post-rehabilitation (4.9, p = 0.001) and 2-month post-rehabilitation (4.9, p = 0.002). The experimental group yielded higher 2-month post-rehabilitation gain than the control (6.0 versus 4.0) but no significance was found (p = 0.475). However, among subjects with positive gain (N = 6 and 7), the initial difference of 2.8 between the two groups was increased to a significant 6.5 (p = 0.019) after adjustment for age and gender. Hence this study provides evidence that BCI-driven robotic rehabilitation is effective in restoring motor control for stroke.
In our every day life, our brain is constantly processing information and paying attention, reacting accordingly, to all sorts of sensory inputs (auditory, visual, etc.). In some cases, there is a need to accurately measure a person's level of attention to monitor a sportsman performance, to detect attention deficit hyperactivity disorder (ADHD) in children, to evaluate the effectiveness of neuro-feedback treatment, etc. In this paper we propose a novel approach to extract, select and learn spectral-spatial patterns from electroencephalogram (EEG) recordings. Our approach improves over prior-art methods that were typically, only concerned with power of specific EEG rhythms from few individual channels. In this new approach, spectral-spatial features from multichannel EEG are extracted by a two filtering stages: a filter-bank (FB) and common spatial patterns (CSP) filters. The most important features are selected by a mutual information (MI) based feature selection procedure and then classified using Fisher linear discriminant (FLD). The outcome is a measure of the attention level. An experimental study was conducted with 5 healthy young male subjects with their EEG recorded in various attention and non-attention conditions (opened eyes, closed eyes, reading, counting, relaxing, etc.). EEGs were used to train and evaluate the model using 4x4fold cross-validation procedure. Results indicate that the new proposed approach outperforms the prior-art methods and can achieve up to 89.4% classification accuracy rate (with an average improvement of up to 16%). We demonstrate its application with a two-players attention-based racing car computer game.
Conventional brain computer interfaces rely on a guided calibration procedure to address the problem of considerable variations in electroencephalography (EEG) across human subjects. This calibration, however, implies inconvenience to the end users. In this paper, we propose an online-adaptive-learning method to address this problem for P300-based brain computer interfaces. By automatically capturing subject-specific EEG characteristics during online operation, this method allows a new user to start operating a P300-based brain-computer interface without guided (supervised) calibration. The basic principle is to first learn a generic model termed subject-independent model offline from EEG of a pool of subjects to capture common P300 characteristics. For a new user, a new model termed subject-specific model is then adapted online based on EEG recorded from the new subject and the corresponding labels predicted by either the subject-independent model or the adapted subject-specific model, depending on a confidence score. To verify the proposed method, a study involving 10 healthy subjects is carried out and positive results are obtained. For instance, after 2-4 min online adaptation (spelling of 10-20 characters), the accuracy of the adapted model converges to that of a fully trained supervised subject-specific model.
This paper describes an initial study of non-invasive electroencephalograph (EEG)-based Brain Computer Interface (BCI) application on Stroke patients. The purpose of this study is to combine BCI and robotic arm for after-stroke rehabilitation exercises. A clinically-proven MANUS robotic rehabilitation shell is integrated with the NeuroComm BCI platform, whereby the robotic control mechanism is complemented by the motor imagery of the patient. 8 hemiparetic stroke patients with varying degrees of paralysis on the unilateral upper extremity are recruited for this study. The results show that most BCI-naive hemiparetic stroke patients are capable of operating the BCI effectively, hence motivates further clinical studies on the extent of how BCI-based robotic rehabilitation are comparable with the control group that uses only robotic rehabilitation.
In overt reading and singing tasks, actual vocalization of words in a rhythmic fashion is performed. During execution of these tasks, the role of underlying vascular processes in relation to cortical excitability changes in a spatial manner is uncertain. Our objective was to investigate cortical excitability changes during reading and singing with transcranial magnetic stimulation (TMS), as well as vascular changes with nearinfrared spectroscopy (NIRS). Findings with TMS and NIRS were correlated. TMS and NIRS recordings were performed in 5 normal subjects while they performed reading and singing tasks separately. TMS was applied over the left motor cortex at 9 positions 2.5 cm apart. NIRS recordings were made over these identical positions. Although both TMS and NIRS showed significant mean cortical excitability and hemodynamic changes from baseline during vocalization tasks, there was no significant spatial correlation of these changes evaluated with the 2 techniques over the left motor cortex. Our findings suggest that increased left-sided cortical excitability from overt vocalization tasks in the corresponding "hand area" were the result of "functional connectivity," rather than an underlying "vascular overflow mechanism" from the adjacent speech processing or face/mouth areas. Our findings also imply that functional neurophysiological and vascular methods may evaluate separate underlying processes, although subjects performed identical vocalization tasks. Future research combining similar methodologies should embrace this aspect and harness their separate capabilities.
This paper presents a subject-independent EEG (Electroencephalogram) classification technique and its application to a P300-based word speller. Due to EEG variations across subjects, a user calibration procedure is usually required to build a subject-specific classification model (SSCM). We remove the user calibration through the boosting of a committee of weak classifiers learned from EEG of a pool of subjects. In particular, we ensemble the weak classifiers based on their confidence that is evaluated according to the classification consistency. Experiments over ten subjects show that the proposed technique greatly outperforms the supervised classification models, hence making P300-based BCIs more convenient for practical uses.
Cognitive processes, such as motor intention, attention, and higher level motivational states are important factors that govern motor performance and learning. Current robot-assisted rehabilitative programs focus only on the physical aspects of training. In this paper, we propose a framework for motor rehabilitation based on the augmentation of cognitive channels of patient-robot interactions and using it to deliver a more optimal therapy. By examining the cognitive processes involved in motor control and adaptation, it is argued that optimal therapy needs to be considered in the context of a complete motor scheme consisting not only of sensorimotor signals, but also their interactions with cognitive operations, such as motor planning, attention, and motivation, which mediate motor learning. We outline a few BCI-based modules for the detection and monitoring of relevant cognitive processes, which provide inputs for the robot to automatically modulate parameters of the rehabilitation protocol. Preliminary investigations on a BCI module for detection of motor intention, performed on a small group of stroke patients, show feasible accuracies.
In this paper, an electroencephalogram (EEG)-based brain computer interface (BCI) is proposed for two dimensional cursor control. The horizontal and vertical movements of the cursor are controlled by mu/beta rhythm and P300 potential respectively. The main advantages of this system are: (i) two almost independent control signals are produced simultaneously; (ii) the cursor can be moved from a random position to another random position in a screen. These advantages have been demonstrated in our experiment and data analysis.
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.
Asynchronous control is an important issue for brain-computer interfaces (BCIs) working in real-life settings, where the machine should determine from brain signals not only the desired command but also when the user wants to input it. In this paper, we propose a novel computational approach for robust asynchronous control using electroencephalogram (EEG) and a P300-based oddball paradigm. In this approach, we first address the mathematical modeling of target P300, nontarget P300, and noncontrol signals, by using Gaussian distribution models in a support vector margin space. Furthermore, we derive a method to compute the likelihood of control state in a time window of EEG. Finally, we devise a recursive algorithm to detect control states in ongoing EEG for online application. We conducted experiments with four subjects to study both the asynchronous BCI's receiver operating characteristics and its performance in actual online tests. The results show that the BCI is able to achieve an averaged information transfer rate of approximately 20 b/min at a low false positive rate (one event per minute).
This paper proposes an approach to learn subject- independent P300 models for EEG-based brain-computer inter- faces. The P300 models are first learned using a pool of existing subjects and Fisher linear discriminant, and then autonomously adapted to the unlabeled data of a new subject using an unsupervised machine learning technique. In data analysis, we apply this technique to a set of EEG data of 10 subjects performing word spelling in an oddball paradigm. The results are very positive: the adapted models with unlabeled data yield virtually the same classification accuracy as the conventional methods with labeled data. Therefore, it proves the feasibility of P300-based BCIs which can be applied directly to a new subject without training sessions.
Cerebrovascular infarctions and force traumas are major causes of damage to the central nervous system and result in paralysis. As the central nervous system has limited ability to regenerate or repair itself, regaining of voluntary motor function is thus highly dependent on neuroplasticity, the brain's ability to reorganize to compensate for lost function or new environmental requirements. Neuroplasticity, unfortunately, seldom allows paralysed persons to recover fully, and there is a need to reanimate the paretic limb. This paper presents a neuroprosthesis that employs a brain-computer interface (BCI) to control functional electrical stimulation (FES) of skeletal muscles to move the arm. The EEG-based BCI analyses μ (mu) and β (beta) frequency bands and is used to control a cursor on-screen. The GUI allows the user to perform either of two movements through electrical stimulation of the flexors or extensors. When the user's intention is deciphered, electrical pulses are sent via surface electrodes to the respective muscles to produce tetanic contraction. Joint angle is used as feedback to control stimulation of the limb in order to achieve the desired movement.
This paper presents an unsupervised subject modeling technique and its application to a P300-based word speller. Due to EEG variations across subjects, a special training procedure is required to learn a subject-specific classification model (SSCM). To deal with the inter-subject variation, we first study a subject independent classification model (SICM) that is learned from EEG of a pool of subjects. Next we further adapt the SICM by learning from a subset of the pooled EEG that is automatically selected based on its similarity to the EEG of a new subject. Experiments over ten healthy subjects show that the SICM learned from all pooled EEG outperforms the cross-subject models greatly. More importantly, the adapted SICM achieves virtually the same performance as the SSCM, hence removing the complicated and tedious training procedure.
This clinical study investigates whether the spatial patterns of hemiparetic stroke patients operating a non-invasive Motor Imagery-based Brain Computer Interface (MI-BCI) is comparable to healthy subjects. The spatial patterns for a specific frequency range are generated using the common spatial pattern (CSP) algorithm, of which is highly successful for discriminating two classes of EEG measurements in MI-BCI. The spatial patterns illustrate how the presumed sources project on the scalp and are effective in verifying the neurophysiological plausibility of the computed solution. The spatial patterns show focused activity in ipsilateral as well as contralateral hemisphere with respect to the hand by tapping or motor imagery in 2 BCI-artful healthy subjects and 12 BCI-naïve hemiparetic stroke patients. The results also show that neurophysiologically interpretable spatial patterns is more common in performing motor imagery compared to finger tapping by hemiparetic stroke patients. Hence, this shows that hemiparetic stroke patients are capable of operating MI-BCI.
This clinical study investigates whether the performance of hemiparetic stroke patients operating a non-invasive Motor Imagery-based Brain-Computer Interface (MI-BCI) is comparable to healthy subjects. The study is performed on 8 healthy subjects and 35 BCI-naïve hemiparetic stroke patients. This study also investigates whether the performance of the stroke patients in operating MI-BCI correlates with the extent of neurological disability. The performance is objectively computed from the 10 x 10-fold cross-validation accuracy of employing the Filter Bank Common Spatial Pattern (FBCSP) algorithm on their EEG measurements. The neurological disability is subjectively estimated using the Fugl-Meyer Assessment (FMA) of the upper extremity. The results show that the performance of BCI-naïve hemiparetic stroke patients is comparable to healthy subjects, and no correlation is found between the accuracy of their performance and their motor impairment in terms of FMA.
- Aug 2007
- Multimedia and Expo, 2007 IEEE International Conference on
The "Brainy Communicator" is a novel state-of-the-art Brain-Computer Interface (BCI) system developed by the Institute for Infocomm Research, Singapore. It allows users to interact with the environment by using just the brain signals. The primary objective of this work is to enhance the quality of life for the people with severe disabilities, by helping them regain some control and communication abilities.
- Jun 2007
- Neural Engineering, 2007. CNE '07. 3rd International IEEE/EMBS Conference on
The motor imagery detection is a very important problem in the asynchronous control for direct brain computer interface. To address this issue, this paper proposes a novel detection method based on subband entropy analysis in a selected frequency band. The basic idea of this method is that, in some specific frequency band, the complexity (or randomness) of brain signal during the stage of concentrating on the motor imagery is lower than that of free thinking. Once the optimal frequency band is selected, the subband entropy $an indicator of complexity and randomness - can be used for detecting the motor imagery. In this work, we develop the method using only one dipolar EEG channel. Furthermore, we propose a system calibration method based on an empirical measurement what we refer as unsupervised discriminative index (UDI). The proposed calibration method is rapid and able to avoid a typical problem of asynchronous BCI training that is the correct labeling of continuous EEG signal. The proposed method not only improve the accuracy of the detection but free from parameter tweaking. The experiment conducted on three different subjects shows advantage of the proposed method over the conventional framework based on fixed-band filter and energy feature. A detection accuracy up to 77% at false positive rate of 2% was obtained without any subject training.
- Apr 2007
Amyotrophic lateral sclerosis, or ALS, is a degenerative disease of the motor neurons that eventually leads to complete paralysis. We are developing a wheelchair system that can help ALS patients, and others who can't use physical interfaces such as joysticks or gaze tracking, regain some autonomy. The system must be usable in hospitals and homes with minimal infrastructure modification. It must be safe and relatively low cost and must provide optimal interaction between the user and the wheelchair within the constraints of the brain-computer interface. To this end, we have built the first working prototype of a brain-controlled wheelchair that can navigate inside a typical office or hospital environment. This article describes the BCW, our control strategy, and the system's performance in a typical building environment. This brain-controlled wheelchair prototype uses a P300 EEG signal and a motion guidance strategy to navigate in a building safely and efficiently without complex sensors or sensor processing
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.
Effective spatial filtering plays a key role in motor imagery classification. This paper presents a novel approach to spatial filtering of EEG signal by modelling time-variant spatial patterns. This is in contrast to conventional Common Spatial Pattern which assumes static spatial patterns in a motor imagery trial. We define the model such that it accounts for relatively higher order dynamics in EEG. Furthermore, we formulate the training of the model as a dual optimization problem, and we derive an iterative optimization algorithm using quadratically constrained quadratic programming. Our experimental results on healthy subjects indicates that the proposed method is able to produce higher classification accuracy.
NeuroComm is a platform to develop real time Brain Computer Interface (BCI) applications. This paper introduces the basic modules of this platform and discusses some implementation issues. With a user management module, our system is user friendly and suitable for multiple users. Also, with flexible configuration files and signal processing algorithm libraries, it is easier to integrate multiple BCI applications into one system. The NeuroComm platform also acts as a flexible tool for BCI research.
Asynchronous control is a critical issue in developing brain-computer interfaces for real-life applications, where the machine should be able to detect the occurrence of a mental command. In this paper we propose a computational approach for robust asynchronous control using the P300 signal, in a variant of oddball paradigm. First, we use Gaussian models in the support vector margin space to describe various types of EEG signals that are present in an asynchronous P300-based BCI. This allows us to derive a probability measure of control state given EEG observations. Second, we devise a recursive algorithm to detect and locate control states in ongoing EEG. Experimental results indicate that our system allows information transfer at approx. 20bit/min at low false alarm rate (1/min).
In this paper we present a new scheme for brain signal processing and classification for electroencephalogram based brain-computer interfaces, by emphasizing the extraction of space-time-frequency feature as well as the combination of classifiers. In particular, we use wavelet packets as a time-frequency analysis tool and employ sparse component analysis to recover source components in the brain signals. We subsequently apply multi-class common spatial pattern filters to the signals and thus obtain important space-time-frequency features for discrimination. Furthermore, a Bayesian method is developed to boost the system, by combining multiple support vector machines in a probabilistic way. We have tested the proposed scheme on real multi-class motor imagery signals, and its efficacy has been demonstrated
This paper presents the first working prototype of a brain controlled wheelchair able to navigate inside a typical office or hospital environment. This brain controlled wheelchair (BCW) is based on a slow but safe P300 interface. To circumvent the problem caused by the low information rate of the EEG signal, we propose a motion guidance strategy providing safe and efficient control without complex sensors or sensor processing. Experiments demonstrated that healthy subjects could safely control the wheelchair in an office like environment, without any training
- Feb 2006
This paper attempts to make use of brain computer interface (BCI) in implementing an application called the media communication center for the paralyzed people. The application is based on the event-related potential called P300 to perform button selections on media and communication programs such as the mp3 player, video player, photo gallery and e-book. One of the key issues in such system is the usability. We study how various tasks affect the application operation, in particular, how typical mental activities cause false trigger during the operation of the application. We study the false acceptance rate under the conditions of closing eyes, reading a book, listening to music and watching a video. Data from 5 subjects is used to obtain the false rejection rate and false acceptance rate of the BCI system. Our study shows that different mental activities show different impacts on the false acceptance performances.
This paper presents a novel method for signal localization for building high-performance brain-computer interfaces using near-infrared spectroscopy. It first proposes a kernel-based model to represent haemodynamic signals of interest under parameterized transformations. A mathematical solution is therefore derived to locate the signals by estimating the parameters. We employ a support vector machine to classify the located signals into left/right hand movements. We evaluate the method on both simulated and real world data, with positive results suggesting the method's high efficacy. This work can be extended to other systems using e.g. fMRI and EEG
This paper presents a new visual tracking method that can achieve accurate estimation of affine transformation and precise spatial-color representation. The estimation of transformation provides more information than translation for better motion understanding and also helps maintain the precise representation; the precise representation enables tracking objects in highly-cluttered environment. The basis of the method is a kernel-based similarity measure called affine matching that describes the relationship between image regions with respect to affine transformation parameters. Based on the similarity measure, a mathematical solution is derived for estimating the transformation parameters for moving objects in videos. Various experiments have yielded positive results.
This paper presents a novel approach to improving the robustness of brain-computer interfaces by using a statistical model of brain signals especially P300. We study the distributions of support vector machine scores for the signals and derive a posteriori probability model of P300/non-P300. We further derive a statistical model for multi-trial brain signals, and apply it to the rejection of undesired signals. Six subjects have been involved in an experimental study. The results demonstrate that the P300 model and the rejection method are appropriate and can help improve the robustness of the system significantly.