Xiaomu Song

Improving brain-computer interface classification using adaptive common spatial patterns

Common Spatial Patterns (CSP) is a widely used spatial filtering technique for electroencephalography (EEG)-based brain-computer interface (BCI). It is a two-class supervised technique that needs subject-specific training data. Due to EEG nonstationarity, EEG signal may exhibit significant intra- and inter-subject variation. As a result, spatial filters learned from a subject may not perform well for data acquired from the same subject at a different time or from other subjects performing the same task. Studies have been performed to improve CSPs performance by adding regularization terms into the training. Most of them require target subjects training data with known class labels. In this work, an adaptive CSP (ACSP) method is proposed to analyze single trial EEG data from single and multiple subjects. The method does not estimate target datas class labels during the adaptive learning and updates spatial filters for both classes simultaneously. The proposed method was evaluated based on a comparison study with the classic CSP and several CSP-based adaptive methods using motor imagery EEG data from BCI competitions. Experimental results indicate that the proposed method can improve the classification performance as compared to the other methods. For circumstances where true class labels of target data are not instantly available, it was examined if adding classified target data to training data would improve the ACSP learning. Experimental results show that it would be better to exclude them from the training data. The proposed ACSP method can be performed in real-time and is potentially applicable to various EEG-based BCI applications.

A SVM-based quantitative fMRI method for resting-state functional network detection

Resting-state functional magnetic resonance imaging (fMRI) aims to measure baseline neuronal connectivity independent of specific functional tasks and to capture changes in the connectivity due to neurological diseases. Most existing network detection methods rely on a fixed threshold to identify functionally connected voxels under the resting state. Due to fMRI non-stationarity, the threshold cannot adapt to variation of data characteristics across sessions and subjects, and generates unreliable mapping results. In this study, a new method is presented for resting-state fMRI data analysis. Specifically, the resting-state network mapping is formulated as an outlier detection process that is implemented using one-class support vector machine (SVM). The results are refined by using a spatial-feature domain prototype selection method and two-class SVM reclassification. The final decision on each voxel is made by comparing its probabilities of functionally connected and unconnected instead of a threshold. Multiple features for resting-state analysis were extracted and examined using an SVM-based feature selection method, and the most representative features were identified. The proposed method was evaluated using synthetic and experimental fMRI data. A comparison study was also performed with independent component analysis (ICA) and correlation analysis. The experimental results show that the proposed method can provide comparable or better network detection performance than ICA and correlation analysis. The method is potentially applicable to various resting-state quantitative fMRI studies.

A study of kernel CSP-based motor imagery brain computer interface classification

The Common Spatial Patterns (CSP) method is a widely used spatial filtering technique that can extract discriminative features for Electroencephalogram (EEG)-based brain computer interface (BCI) classification tasks. Since the EEG signal acquired on the scalp is a nonlinear composition of multiple signal and noise sources, in order to characterize the nonlinear data structure, nonlinear CSP methods have been proposed by using the kernel technique. Most kernel CSP methods calculate temporal covariance structure in a kernel feature space that leads to a large kernel matrix with each dimension equal to the number of time points multiplied by the number of classes. In this work, a kernel CSP method exploiting spatial covariance structure in the feature space is developed where the size of kernel matrix is the number of EEG channels, which is usually much less than that of time points. The proposed method was evaluated using motor imagery EEG data. Results indicate that the kernel CSP using spatial analysis can provide comparable performance to the existing methods using temporal analysis with less computational load.

Data-Driven and Predefined ROI-Based Quantification of Long-Term Resting-State fMRI Reproducibility.

Resting-state functional magnetic resonance imaging (fMRI) is a promising tool for neuroscience and clinical studies. However, there exist significant variations in strength and spatial extent of resting-state functional connectivity over repeated sessions in a single or multiple subjects with identical experimental conditions. Reproducibility studies have been conducted for resting-state fMRI where the reproducibility was usually evaluated in predefined regions-of-interest (ROIs). It was possible that reproducibility measures strongly depended on the ROI definition. In this work, this issue was investigated by comparing data-driven and predefined ROI-based quantification of reproducibility. In the data-driven analysis, the reproducibility was quantified using functionally connected voxels detected by a support vector machine (SVM)-based technique. In the predefined ROI-based analysis, all voxels in the predefined ROIs were included when estimating the reproducibility. Experimental results show that (1) a moderate to substantial within-subject reproducibility and a reasonable between-subject reproducibility can be obtained using functionally connected voxels identified by the SVM-based technique; (2) in the predefined ROI-based analysis, an increase in ROI size does not always result in higher reproducibility measures; (3) ROI pairs with high connectivity strength have a higher chance to exhibit high reproducibility; (4) ROI pairs with high reproducibility do not necessarily have high connectivity strength; (5) the reproducibility measured from the identified functionally connected voxels is generally higher than that measured from all voxels in predefined ROIs with typical sizes. The findings (2) and (5) suggest that conventional ROI-based analyses would underestimate the resting-state fMRI reproducibility.

A unified machine learning method for task-related and resting state fMRI data analysis

Functional magnetic resonance imaging (fMRI) aims to localize task-related brain activation or resting-state functional connectivity. Most existing fMRI data analysis techniques rely on fixed thresholds to identify active voxels under a task condition or functionally connected voxels in the resting state. Due to fMRI non-stationarity, a fixed threshold cannot adapt to intra- and inter-subject variation and provide a reliable mapping of brain function. In this work, a machine learning method is proposed for a unified analysis of both task-related and resting state fMRI data. Specifically, the mapping of brain function in a task condition or resting state is formulated as an outlier detection process. Support vector machines are used to provide an initial mapping and refine mapping results. The method does not require a fixed threshold for the final decision, and can adapt to fMRI non-stationarity. The proposed method was evaluated using experimental data acquired from multiple human subjects. The results indicate that the proposed method can provide reliable mapping of brain function, and is applicable to various quantitative fMRI studies.

Spatially regularized machine learning for task and resting-state fMRI

BACKGROUNDnReliable mapping of brain function across sessions and/or subjects in task- and resting-state has been a critical challenge for quantitative fMRI studies although it has been intensively addressed in the past decades.nnnNEW METHODnA spatially regularized support vector machine (SVM) technique was developed for the reliable brain mapping in task- and resting-state. Unlike most existing SVM-based brain mapping techniques, which implement supervised classifications of specific brain functional states or disorders, the proposed method performs a semi-supervised classification for the general brain function mapping where spatial correlation of fMRI is integrated into the SVM learning. The method can adapt to intra- and inter-subject variations induced by fMRI nonstationarity, and identify a true boundary between active and inactive voxels, or between functionally connected and unconnected voxels in a feature space.nnnRESULTSnThe method was evaluated using synthetic and experimental data at the individual and group level. Multiple features were evaluated in terms of their contributions to the spatially regularized SVM learning. Reliable mapping results in both task- and resting-state were obtained from individual subjects and at the group level.nnnCOMPARISON WITH EXISTING METHODSnA comparison study was performed with independent component analysis, general linear model, and correlation analysis methods. Experimental results indicate that the proposed method can provide a better or comparable mapping performance at the individual and group level.nnnCONCLUSIONSnThe proposed method can provide accurate and reliable mapping of brain function in task- and resting-state, and is applicable to a variety of quantitative fMRI studies.

Incremental versus non-incremental learning in adaptive common spatial patterns

Extracting reliable and discriminative features remains a critical challenge in the development of brain computer interface (BCI) techniques. Common spatial patterns (CSP) is frequently used for spatial filtering and feature extraction in electroencephalography (EEG)-based BCI. It performs a supervised and subject-specific learning of EEG data acquired in two different task conditions. Incremental learning has been used in CSP to adapt to a target subjects data by including classified data in training data and re-estimating spatial filters. In practical circumstances where no user feedback is instantly available to provide true class labels of target trials, misclassified EEG trials will be added to the training data of a wrong class, and potentially influence the training of spatial filters and feature extraction. In this study, incremental and non-incremental learning were investigated based on a recently developed adaptive CSP (ACSP) method using multi-subject EEG data. Their performances were compared in terms of intra- and inter-subject classification performances. Experimental results indicate that the non-incremental learning is a better option when true class labels of target data are not provided.

A Kernel Machine-based fMRI Physiological Noise Removal Method

Functional magnetic resonance imaging (fMRI) technique with blood oxygenation level dependent (BOLD) contrast is a powerful tool for noninvasive mapping of brain function under task and resting states. The removal of cardiac- and respiration-induced physiological noise in fMRI data has been a significant challenge as fMRI studies seek to achieve higher spatial resolutions and characterize more subtle neuronal changes. The low temporal sampling rate of most multi-slice fMRI experiments often causes aliasing of physiological noise into the frequency range of BOLD activation signal. In addition, changes of heartbeat and respiration patterns also generate physiological fluctuations that have similar frequencies with BOLD activation. Most existing physiological noise-removal methods either place restrictive limitations on image acquisition or utilize filtering or regression based post-processing algorithms, which cannot distinguish the frequency-overlapping BOLD activation and the physiological noise. In this work, we address the challenge of physiological noise removal via the kernel machine technique, where a nonlinear kernel machine technique, kernel principal component analysis, is used with a specifically identified kernel function to differentiate BOLD signal from the physiological noise of the frequency. The proposed method was evaluated in human fMRI data acquired from multiple task-related and resting state fMRI experiments. A comparison study was also performed with an existing adaptive filtering method. The results indicate that the proposed method can effectively identify and reduce the physiological noise in fMRI data. The comparison study shows that the proposed method can provide comparable or better noise removal performance than the adaptive filtering approach.

A quadcopter controlled by brain concentration and eye blink

Brain computer interface (BCI) is a technology that enables a user to interact with the outside world by measuring and analysing signals associated with neural activity, and mapping an identified neural activity pattern to a behavior or action. In this work, an BCI system was developed where the operation of a quadcopter is controlled by identified brain concentration and eye blink patterns. A portable electroencephalography (EEG) headset is used to acquire neural signal around forehead and both eyes. Acquired EEG data are sent to a data processing computer wirelessly and processed in real-time. Identified brain concentration and eye blink patterns are associated with quadcopter operation commands and transmitted to the remote control that is modified to interface with the computer. The BCI system was evaluated by an experiment study and classification accuracy was calculated. Experimental results indicate that the system can achieve the expected performance without using EEG data from all channels and complicated data processing algorithms.

Brain functional mapping using spatially regularized support vector machines

Quantitative functional magnetic resonance imaging (fMRI) requires reliable mapping of brain function in task-or resting-state. In this work, a spatially regularized support vector machine (SVM)-based technique was proposed for brain functional mapping of individual subjects and at the group level. Unlike most SVM-based fMRI data analysis approaches that conduct supervised classifications of brain functional states or disorders, the proposed technique performs a semi-supervised learning to provide a general mapping of brain function in task-or resting-state. The method can adapt to between-session and between-subject variations of fMRI data, and provide a reliable mapping of brain function. The proposed method was evaluated using synthetic and experimental data. A comparison with independent component analysis methods was also performed using the experimental data. Experimental results indicate that the proposed method can provide a reliable mapping of brain function and be used for different quantitative fMRI studies.