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.

Abstract-One of the most popular feature extraction algorithms for Brain-Computer Interfaces (BCI) is Common Spatial Patterns (CSP). Despite its known efficiency and widespread use, CSP is also known to be very sensitive to noise and prone to overfitting. To address this issue, it has been recently proposed to regularize CSP. In this paper, we present a simple and unifying theoretical framework to design such a Regularized CSP (RCSP). We then present a review of existing RCSP algorithms, and describe how to cast them in this framework. We also propose 4 new RCSP algorithms. Finally, we compare the performances of 11 different RCSP (including the 4 new ones and the original CSP), on EEG data from 17 subjects, from BCI competition data sets. Results showed that the best RCSP methods can outperform CSP by nearly 10% in median classification accuracy and lead to more neurophysiologically relevant spatial filters. They also enable us to perform efficient subject-to-subject transfer. Overall, the best RCSP algorithms were CSP with Tikhonov Regularization and Weighted Tikhonov Regularization, both proposed in this paper.

Recently, artificial intelligence and machine learning in general have demonstrated remarkable performances in many tasks, from image processing to natural language processing, especially with the advent of deep learning (DL). Along with research progress, they have encroached upon many different fields and disciplines. Some of them require high level of accountability and thus transparency, for example, the medical sector. Explanations for machine decisions and predictions are thus needed to justify their reliability. This requires greater interpretability, which often means we need to understand the mechanism underlying the algorithms. Unfortunately, the blackbox nature of the DL is still unresolved, and many machine decisions are still poorly understood. We provide a review on interpretabilities suggested by different research works and categorize them. The different categories show different dimensions in interpretability research, from approaches that provide "obviously" interpretable information to the studies of complex patterns. By applying the same categorization to interpretability in medical research, it is hoped that: 1) clinicians and practitioners can subsequently approach these methods with caution; 2) insight into interpretability will be born with more considerations for medical practices; and 3) initiatives to push forward data-based, mathematically grounded, and technically grounded medical education are encouraged.

No abstract

Electroencephalography (EEG)-based motor imagery (MI) brain-computer interface (BCI) technology has the potential to restore motor function by inducing activity-dependent brain plasticity. The purpose of this study was to investigate the efficacy of an EEG-based MI BCI system coupled with MIT-Manus shoulder-elbow robotic feedback (BCI-Manus) for subjects with chronic stroke with upper-limb hemiparesis. In this single-blind, randomized trial, 26 hemiplegic subjects (Fugl-Meyer Assessment of Motor Recovery After Stroke [FMMA] score, 4-40; 16 men; mean age, 51.4 years; mean stroke duration, 297.4 days), prescreened with the ability to use the MI BCI, were randomly allocated to BCI-Manus or Manus therapy, lasting 18 hours over 4 weeks. Efficacy was measured using upper-extremity FMMA scores at weeks 0, 2, 4 and 12. ElEG data from subjects allocated to BCI-Manus were quantified using the revised brain symmetry index (rBSI) and analyzed for correlation with the improvements in FMMA score. Eleven and 15 subjects underwent BCI-Manus and Manus therapy, respectively. One subject in the Manus group dropped out. Mean total FMMA scores at weeks 0, 2, 4, and 12 weeks improved for both groups: 26.3±10.3, 27.4±12.0, 30.8±13.8, and 31.5±13.5 for BCI-Manus and 26.6±18.9, 29.9±20.6, 32.9±21.4, and 33.9±20.2 for Manus, with no intergroup differences (P=.51). More subjects attained further gains in FMMA scores at week 12 from BCI-Manus (7 of 11 [63.6%]) than Manus (5 of 14 [35.7%]). A negative correlation was found between the rBSI and FMMA score improvement (P=.044). BCI-Manus therapy was well tolerated and not associated with adverse events. In conclusion, BCI-Manus therapy is effective and safe for arm rehabilitation after severe poststroke hemiparesis. Motor gains were comparable to those attained with intensive robotic therapy (1,040 repetitions/session) despite reduced arm exercise repetitions using EEG-based MI-triggered robotic feedback (136 repetitions/session). The correlation of rBSI with motor improvements suggests that the rBSI can be used as a prognostic measure for BCI-based stroke rehabilitation.